Telecommunication operators in India have faced significant challenges due to the high rate of customer churn, which is a consequence of customer dissatisfaction and fierce competition in the market. This paper is a new machine learning approach that is committed to forecasting customer churn and optimizing Internet Service Provider (ISP) plans. The suggested system merges customers’ behavioral, billing data, and service quality indicators and uses XGBoost for churn prediction. The framework substantially differs from common recommendation systems in that it assesses the network performance information of the user, in real-time, and thereby matches plan recommendations with it. Besides, a feature importance analysis is carried out to determine the main factors leading to customer churn. To maintain robustness and generalizability, the model is tested on multiple metrics of performance. The proposed framework, through experiments, has been able to prove that it brings about greater predictive accuracy, user satisfaction, and customer retention.